HAR_IMU_Stretch

GitHub repo for the iSPL IMU-Stretch dataset and implementation code of the paper "An Investigation on Deep Learning-Based Human Activity Recognition Using IMUs and Stretch Sensors" accepted by the ICAIIC2022 conference.

1. iSPL IMU-Stretch Dataset

The iSPL IMU-Stretch dataset was collected from one subject wearing 3 IMUs (at the right wrist, waist and ankle) and 2 fabric stretch sensors (at the knees). 9 different activities are collected: walking, standing, sitting, lying, running, jumping, sit-up, push-up and dancing.

• From each IMU sensor, 3D acceleration, 3D angular velocity and 3D linear acceleration are collected.
• From each stretch sensor, the stretch degree (actually is the capacity of the stretch capacitor) is collected (the data is 1D).
• The data is already pre-processed (interpolation & windowing) and randomly split into train and test sets with a proportion of 70/30 and saved as mat files. Each mat file contains:
  • Window data of 3 IMUs: X_IMUs_train (X_IMUs_test): 4-D array: [number of windows; window size; ]
  • Window data of 2 stretch sensors: X_Stretch_train (X_Stretch_test)
  • Ground truth label: y_train (y_test)

2. w-HAR dataset

• The w-HAR dataset was introduced in the "Bhat, G.; Tran, N.; Shill, H.; Ogras, U.Y. w-HAR: An Activity Recognition Dataset and Framework Using Low-Power Wearable Devices. Sensors 2020, 20, 5356." and is available at the GitHub repo: https://github.com/gmbhat/human-activity-recognition

• The preprocessing process is implemented using MATLAB, the codes are in the folder data/w-har_data

  • Step 0: Import the csv data and save them into mat files for smaller data size and a faster data loading process
  • Step 1: Save the data of each (user-scenario-trial) as a cell in the cell array
    • m_step1_data_combining_motion.m
    • m_step1_data_combining_stretch.m
    • m_step1b_check_frequency.m
  • Step 2: Resampling the data to a new sampling rate
    • m_step2_resampling_v2.m
  • Step 3: Windowing (split the data into a fixed length window with an overlap between adjacent windows)
    • m_step3_windowing.m

• The train/test split process is implemented using python, the codes are in the file split_data.py

3. HAR Models Implementation

• The deep learning models are in the file model_zoo.py
• The training and evaluation tasks are in the files:
  • imu_stretch_ispl.py
  • imu_stretch_wHAR.py

Citation: If you would like to use the iSPL IMU-Stretch Dataset or the materials in this repo, please cite our work:

N. T. H. Thu and D. S. Han, "An Investigation on Deep Learning-Based Activity Recognition Using IMUs and Stretch Sensors," 2022 International Conference on Artificial Intelligence in Information and Communication (ICAIIC), 2022, pp. 377-382, doi: 10.1109/ICAIIC54071.2022.9722621.

If you have any question, please feel free to contact me (thunguyen@knu.ac.kr) or if you find out any error, please let me know by creating a new issue in this repo.

Thank you for your interest in this work!

** Stand on the shoulders of giants **